Study population
Data were derived from the Kailuan study (Trial identification: ChiCTR–TNRC–11001489; Registration number: 11001489), which is a large, prospective, longitudinal and population-based cohort study. Details regarding the design and methods of the study have been previously described [17, 18]. All corporations’ employees, including retirees, between the ages 18 to 98 years were invited to take part in the baseline physical examinations which took place at Kailuan General Hospital and it’s 10 affiliated hospitals between July 2006 and October 2007. Written consent was obtained from 101,510 participants (65.3%) who agreed to take part. All participants were followed up biennially for collection of information regarding potential risk factors and newly diagnosed CRC cases.
In the present study, we excluded 469 subjects with a history of cancer at baseline examination, 3319 subjects who had missing data or unclear results for hepatitis B surface antigen (HBsAg), and 4332 subjects without data of other potential variables including age, sex, body mass index (BMI, in kg/m2), waist circumference (WC, in cm), systolic blood pressure (SBP, in mmHg), diastolic blood pressure (DBP, in mmHg), total cholesterol (TC, in mmol/L), triglyceride, (TG, in mmol/L), high-sensitivity C-reactive protein (hs-CRP, in mg/L), alanine aminotransferase (ALT, in u/L), total bilirubin (TBil in umol/L), diabetes mellitus, family income, educational background, marital status, salt consumption, current smoker, drinking status, physical activity and family history of cancer. A total of 93,390 subjects were finally enrolled in the study, including 74,644 (79.93%) men and 18,746 (20.07%) women. The details of the participants’ screening were shown in Fig. 1. In this current study, participants who were excluded were younger (49.85 ± 14.13 years versus 51.51 ± 12.44 years, P < 0.001), and had lower levels of BMI (24.71 ± 3.45 Kg/m2 versus 25.07 ± 3.49 Kg/m2, P < 0.001), and exhibited lower prevalence of HBV infection (179(2.20%) versus 2598(2.78%), P = 0.0105). This study was in compliance with the Declaration of Helsinki and was approved by the Ethics Review Committee of Kailuan General Hospital and Beijing Shijitan Hospital. Written informed consents were obtained from all participants.
Epidemiological survey and anthropometric parameters
Details of the collection of epidemiological surveys and anthropometric parameters were in accordance with previously published articles [19, 20]. Information on age, sex, lifestyle behaviors, educational background, socioeconomic status, and medical history was recorded via a standard questionnaire which was carried out by the medical staff and trained nurse. BMI was calculated as body weight (kilograms) divided by the square of body height (meters) and categorized into three groups [21]: normal (BMI<24 Kg/m2), overweight (24 ≤ BMI<28 Kg/m2) and obesity (BMI ≥ 28 Kg/m2). Current alcohol consumer was defined as having drunk ≥100 ml/day of alcohol lasting for more than 6 months. Smoking was defined as having 1 cigarette/day at least for more than 6 months. Physical exercise was evaluated from responses regarding the frequency of physical activity (≥3 times/week, ≥30 min/time). Dietary salt intake was self-reported and classified into three categories: low (< 6 g/day), medium (6-9 g/day) or high (≥10 g/day), and high salt diet was defined as consuming salt≥10 g per day on average. Hypertension was defined as: previously diagnosed, and/or a SBP ≥140 mmHg, and/or a DPB ≥90 mmHg, and/or the use of antihypertensive medication [22].
HBV infection and laboratory assessment
Blood samples were collected using vacuum tubes containing EDTA after an overnight fast (≥ 8 h) from each participant at the baseline. The blood was further centrifuged for 10 min at 3000 rotations per minute at 25 °C. Plasma was separated and stored at − 80 °C until laboratory determinations were performed. An auto-analyzer (Hitachi 747; Hitachi, Tokyo, Japan) was used to analyze all the plasma samples at the Kailuan General Hospital central laboratory. HBsAg was detected quantitatively by the enzyme-linked immunosorbent assay (SHANGHAI KEHUA BIO-ENGINEERING, KHB, Shanghai, China) with standard operating procedure. TC and TG were both measured using enzymatic colorimetric method (Mind Bioengineering Co. Ltd., Shanghai, China). Hs-CRP was measured using a high-sensitivity nephelometry assay (Cias Latex CRP-H, Kanto Chemical Co. Inc). ALT was measured using an enzymatic rate method (Mind Bioengineering Co. Ltd., Shanghai, China). Serum TBil was measured using a chemical oxidation method (MedicalSystem Biotechnology, China). Diabetes mellitus was defined as follows [23]: a fasting blood glucose level ≥ 7.0 mmol/L, taking oral hypoglycemic agents or insulin, or a validated physician diagnosis. According to the Guidelines for the Prevention and Treatment of Adult Dyslipidemia in China [24], hypercholesterolemia and hypertriglyceridemia were defined as TC ≥ 6.2 mmol/L and TG ≥ 2.3 mmol/L, respectively. Hyperbilirubinemia was defined as serum TBil> 21 umol/L.
Outcome assessment
The follow-up of each participant began at the end of baseline examination and terminated at the occurrence of any event as follows came first: CRC, death, or end of the follow-up (December 31, 2019). Incident CRC and CRC-unrelated death cases were obtained through routine biennial health examinations. Further CRC and CRC-unrelated death cases were assessed annually by checking discharge summaries from Kailuan General Hospital and its 10 affiliated hospitals where participants received treatments. Furthermore, medical records linked with the Tangshan medical insurance system and death certificates from the Kailuan social security system were checked to further confirm the outcomes. CRC cases were coded according to the International Classification of Diseases, Tenth Revision (ICD–10), and CRC was coded as C18-C20.
Statistical analysis
Person-years of follow-up were calculated from the date of recruitment to the date of incidence, death, or termination of follow-up. Mean (standard deviation, SD) and t-test were used to describe and compare continuous variables. The nonparametric Kruskal-Wallis Test and median (interquartile range) were used to describe and test the differences of skewed distribution variables including serum hs-CRP, ALT, TG and TBil between groups. Proportions and Chi-square tests were used to describe and compare categorical variables. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated by multivariable Cox proportional hazards regression models to evaluate the association between HBV infection and the risk of new-onset CRC. Model 1 was a univariate analysis. Model 2 was adjusted for age (every 10 years) and sex. Model 3 was further adjusted for BMI (normal, overweight, obesity), hypercholesterolemia, hypertriglyceridemia, hs-CRP (< 1 mg/L, 1–3 mg/L, > 3 mg/L), hyperbilirubinemia, elevated alanine aminotransferase, diabetes, family income, educational background, marital status, salt consumption, current smoker, drinking status, physical activity and family history of cancer based on model 2.
Subgroup analyses were performed by sex (women vs. men), age (young group vs. middle-aged group vs. elder group), BMI (normal weight vs. overweight vs. obesity), smoking status (non-smoker vs. smoker) and drinking status (non-drinker vs. drinker). The interactions between HBV infection status and these variables were further tested by multiplicative models.
During follow-up, CRC unrelated death may occur prior to the occurrence of CRC cases. Due to the existence of competing risk events (CRC unrelated death), the observation of new-onset CRC and further interventions can be hindered. Conventional methods for survival analysis such as standard Cox proportional hazards regression may neglect the competing events and overestimate the absolute risk of the disease [25, 26]. Thus, competing risk analysis should be applied to epidemiologic research. In the current study, cause-specific hazards models (CS models) and sub-distribution proportional hazards models (SD models) were used to calculate HRCS and HRSD of CRC with the existence of competing risk by fitting into the standard Cox proportional hazards regressions.
In the sensitivity analysis, we further excluded 39 participants who had CRC within the first year after entry to the cohort to investigate the possibility of reverse causation.
Statistical computations were performed using a commercially available software program (SAS software, version 9.4). Reported P-values are two-sided, and the significance level was set at p < 0.05.